Rule Object Recognition And Grasping Of Indoor Intelligent Mobile Robot

With the rapid development of computer and network technology, the society having brought opportunities for the development of robot technology. In 2015, the robot conference held in Beijing push the robot development to a climax point. As a branch of robot technology, intelligent robot becomes more and more concerned, and becomes the focus of research. This paper combined with the real scene,conducts the research to the rule object recognizing and grasping with the intelligent mobile robot under indoor environment. In the room, the robot identify and locate the specified rule object. In the end, grasping the objects by controlling the movement of the robot. The main work of this paper is summarized as follows:(1) Target detection and recognition. There are some rule objects in the paper, circular, rectangular, and triangular. As for the traditional Hough transform image cannot detect whether there is a circular, rectangular, triangular and others effectively. This paper presents two different solutions of recognition for different rules. The first one is for the circular objects. The segmentation is based on HSV. And then we obtain the binary image, which can be identified by the calculating area and perimeter ratio of the connected domain which contour mark combined with the circular degree.The second one is for rectangular and triangular objects. After the segmentation based on HSV,we proposed using the Harris method to complete the extraction of corners. And analyze the advantages of Harris which does not have the scale invariant features. By improving the algorithm and set the reasonable parameters which can still extract the corner points stably. Because the existence of noises, the number of corners extracted often more than we have predicted. So we should sort the corners which we extracted, and select a certain number of corners, which should be calculated by some forms, just like calculate cosine value of the angle of the corner to recognize them.(2) Camera calibrating. We have researched the algorithm of Zhang Zheng you in depth, and found that the algorithm has some advantages, such as simple operation, high precision, but often not accurate because human factors which could influence the calibratings. In order to obtain the more accurate values, multi-group pictures was captured, and using Matlab toolbox and OpenCV to make sure the calibrating results more accurate.Comparing the setting results and the values which calculated by pinhole imaging principle, we want to verify the correctness of the result.(3) Target location and capture. According to the actual experimental environment, setting the reasonable search strategy, and dividing the equivalence areas for the position of target relative to the robot, the distance algorithm based on the ground plane constraint is proposed, and the error correction is added to improve the positioning accuracy. The Object grabbing, which research on the Pioneer3-AT robot platform and analyze the process of executing the program.And though the location information measured to adjust the robot posture, finally control the movement of the robot.Experiment results show that using the proposed target recognition scheme can identify regular objects accurately, and use the memory less. By computing distance and camera setting results could identify the object localization and then control the robot to grasp the object correctly.